Introducing Saffron Lido Vaults
Introducing Saffron Lido ETH Staking Vaults.
In this article we’ll go over the advancement and evolution of the products we’ve been building over the last year. In our last update we had mentioned our attention focused in parallel on development, auditing, and producing internal testing data. We communicated our efforts to refine our research to find more use cases for Saffron Fixed Income Vaults. Today, we’re announcing a new product based on that research, Saffron Lido Vaults.
Saffron Lido Vaults have a vault structure with two sides: the fixed side and the variable side. All previous products starting with Saffron V1 had the same basic structure, so this should sound familiar.
As evident by the name, our product is built with Lido ETH Staking as the instrument that generates yield. The fixed side enters with principal that will be locked throughout the length of the contract. They receive a 3.27% fixed rate upfront payment from the variable side. The principal deposited by the fixed side is locked on Lido. At the end of the contract, the fixed side reclaims their deposited principal, and the variable side receives all yield generated by the fixed side’s principal while it was earning yield on Lido. This simple yet powerful instrument is one of the cornerstones of traditional finance, and in the first half of 2023, there were 465.9 trillion US dollars in similar contracts.
Our team worked hard to include features that make this an ideal instrument for risk adjustment in personal and institutional portfolios. In the next few sections we will outline these features and explain how they make the Saffron Lido Vaults an unmatched innovation in DeFi.
Freedom to Withdraw
Public feedback we’ve received from testing previous products indicated to us that there is an aversion to having principal locked up for the entire vault duration. This is because of severe downside incurred when there is opportunity cost due to locked funds.
With Saffron Lido Vaults, the fixed side is now able to withdraw all of their principal at any time for a termination fee. This fee is high at the beginning of the vault and decreases over time, nearing zero towards the end of the vault duration.
The variable side can now collect their yield at any time. This improvement to the vault structure, in addition to a new underlying yield source, are major upgrades over the last version of Saffron.
Perpetual Long Integration
In Saffron Lido Vaults we’re planning on integrating a perpetual long platform for users taking the fixed side who want the option to earn more while staking without risking their principal. We consider this strategy to have little downside because its core thesis aligns with ETH staking. Let’s explore this concept more below.
Consider the following scenario: Lido yield is on average 3.5% and you’re staking ETH. If ETH goes down 3.5% on any given day, you’ve effectively lost an entire year of yield in a single day, in terms of USD value. On the other hand, if ETH’s price goes up, then you gain even more because of the yield earned while staking. We’ve decided to integrate perpetuals because anyone staking ETH is also inherently taking a long outlook on ETH.
In the case of liquidation, the amount lost is a small percentage of the total deposit amount because the biggest losses come from simply holding ETH. The depreciation of the principal allocation in ETH is an unavoidable downside of staking ETH anyway. We can reduce the chance of liquidation and use the volatility of the market to our advantage while the market is trending downward by employing more advanced strategies.
Strategies
We present below strategies to select optimal leverage to maximize profit and avoid liquidation.
We use a method called “Walk Forward Optimization” for this. WFO is a general methodology to optimize parameters relevant to a trading strategy based on looking at historical data. In our case we’re going to optimize for selecting the most profitable leverage multiplier. In short, WFO methodology splits the historical dataset into chunks, where each chunk is divided into two parts; training and testing. We use the training part of the data to optimize all relevant parameters to the trading strategy that is being tested. Then, with these parameters, we use the test part of the data to measure the performance of the strategy. We consider this to be one simulation. Then we move on to the next chunk and repeat. In the end, we can review a set of performances for the strategy, which is a better way to validate it than with a single backtesting simulation.
Here we tested strategies to select leverage for a long ETH position. We used the training set for each chunk to find the optimal leverage between 1 and 20x. In these simulations we used a 180-day training set and a 60-day performance test set. Below we show preliminary results for the performance of a few perpetual long strategies. We filtered for simulations where the ETH price increases during the test set (Table 1).
The APR rates here are calculated as gains in ETH.
The strategies named “1.5x”, “2.0x”, and “4.0x” are static strategies, and they represent the respective leverage applied to a long position in the test set independently from the training set. Obviously, the strategy “4.0x” has the largest mean APR (7.22%) among the static strategies.
The strategies named “Stdx2” and “Std” are dynamic strategies. The “Std” strategy calculates one standard deviation of the ETH price in the training set and sets the leverage for the test set to be as large as possible using that price as the liquidation price. The “Stdx2” strategy is almost exactly the same, but instead considers that the price could drop further to two standard deviations below the starting price, leading to a more conservative leverage choice. The “Std” strategy performed the best with the highest average APR of 7.62%.
Performance in the General Case
In cases where the price of ETH decreases, PnL < 0, and losses are amplified by the leverage selection up to the point where all yield is lost due to liquidation. This has a large impact on the overall average performance APR. Obviously, if the price of ETH decreases, the proportional losses in principal far outweighs any losses in earned interest. Keep in mind that the Saffron vault never risks any principal because liquidations are only applied to the fixed premium that is paid up-front to the fixed side.
For completeness, we show the results for all simulations, including those when ETH price decreases (Table 2). The important data here is the percentage of liquidations. The “1.5x” and “Stdx2” strategies were never liquidated. The best-performing strategy saw 22.7% of positions liquidated. We also noticed that the dynamic strategy “Std” has a lower liquidation rate than the best-performing static strategy “4.0x”, highlighting the power of dynamic strategies to determine leverage for long positions.
Thus, strategies can fit different users with unique risk tolerance. In Figure 1 (below) we show the performance of each long position simulated.
Conclusion
This article reviews the design, development, and innovation of Saffron Lido Vaults, a new ETH staking product produced by comprehensive research processes over the past year. By leveraging Lido’s ETH staking protocol, we give users access to more flexible and profitable ETH staking options, addressing user concerns over locked principal and opportunity costs. We’ve also outlined the balance between maximizing profits and minimizing risk. Saffron’s web3 app will give users information to choose which strategy fits them best, while also providing them with the ability to execute it on-chain.
Looking briefly to the future–after we’ve launched Saffron Lido Vaults–we’ll be bringing this instrument to other L1 single-staking opportunities. In the meantime we’re more than happy to take feedback and answer any questions you might have about Saffron Lido Vaults in our Telegram or Discord.
Community resources
Web3 app: https://saffron.finance
Governance: https://gov.saffron.finance
Telegram: https://t.me/saffronfinance
Discord: https://discord.gg/pDXpXKY
Twitter: https://twitter.com/saffronfinance_
Github: https://github.com/saffron-finance/saffron
